Federated Learning with Model Personalization: A Survey

Federated Learning (FL) enables multiple clients to train models collaboratively without sharing local data, but a single global model often struggles with data heterogeneity. Personalized Federated Learning (PFL) addresses this by tailoring models to individual clients while preserving FL’s benefits. This survey categorizes PFL techniques into client-specific models, meta-learning, clustered FL, and multi-task learning. It examines the balance between personalization and generalization, along with challenges like privacy, communication efficiency, and fairness. We also explore real-world applications in healthcare, finance, and IoT, highlighting the need for adaptive strategies. Lastly, we discuss open research directions in privacy-preserving personalization, adaptive optimization, and hybrid FL frameworks, offering insights for researchers and practitioners.